The rankings tell one story. In the Global AI Index, Finland sits eleventh, Denmark sixteenth, Sweden seventeenth, and Norway twenty-fifth: all four inside the top twenty-five, which is respectable for countries whose combined population is smaller than Poland's. The trajectory tells another story. A decade ago the Nordic countries stood at the frontier of digital governance, moving first on electronic government, digital identity, open data, and public-sector innovation. The rest of the world studied their models. The rest of the world is now building AI at a scale and pace that the Nordics, for all their institutional advantages, cannot match.
In June 2025 the Nordic Council of Ministers approved 30 million Danish kroner, approximately four million euros, to establish a Nordic-Baltic AI centre called New Nordics AI, bringing together AI Sweden, AI Finland, Digital Dogme of Denmark, IKT Norge, and Iceland's Almannarómur. Five countries, five organisations, one centre. The ministers of digitalisation from all five nations declared that the Nordic region must be "Europe's driving force in the AI race".1
Four million euros. For context: the European Commission has pledged twenty billion euros for AI Giga-Factories. The Berlin summit on European digital sovereignty produced twelve billion euros in voluntary corporate commitments. The Stargate Project in the United States promised five hundred billion dollars. DeepSeek trained a frontier model for six million. The entire three-year budget of the Nordic centre is less than the cost of training a single competitive large language model.
The paradox is not that the Nordic countries lack ambition. It is that they possess, in abundance, the qualities every AI governance framework identifies as essential—digital literacy, institutional trust, a data protection culture, transparent government, high broadband penetration, educated workforces—and that these qualities have not translated into competitive AI capability. The foundations are impeccable. The structure built on them is modest.
What follows is an account of what the Nordic countries got right, what they lack, and whether pooling sovereignty and governance quality can compensate for scale and capital.
What the Nordics got right
The Nordic countries were digital-first before the term existed. Finland published its first national AI strategy in 2017, among the earliest in Europe.2 Estonia, a Baltic neighbour with deep Nordic institutional ties, pioneered electronic residency and digital government infrastructure that became a model for the continent. Denmark's digital-first government strategy, launched in 2011 and updated continuously since, made digital interaction the default channel for public services. Sweden invested heavily in broadband, achieving near-universal high-speed connectivity.
The institutional foundations are equally strong. Public trust in government is among the highest in the world, which is a precondition for the data-sharing and digital identity systems that AI applications require.3 A data protection culture, built on decades of privacy regulation that predated and influenced the General Data Protection Regulation, means citizens are accustomed to transparent data practices and institutions have developed sophisticated frameworks for handling personal information. Digital literacy rates are consistently among the highest globally, supported by education systems that have integrated technology into curricula for decades.
These are not incidental advantages. They are the precise qualities that every major AI governance framework identifies as necessary for responsible development and deployment, whether the EU AI Act, the principles published by the Organisation for Economic Co-operation and Development, or the UNESCO Recommendation on AI Ethics. Representative datasets require trusted data governance. Human oversight requires digitally literate populations. Transparency requires institutions that citizens believe in. The Nordic countries have all of these, and had built the preconditions for responsible AI before most countries recognised the need.
The research capacity is genuine. Finland hosts LUMI, one of the world's most powerful supercomputers, operated in Kajaani by the European high-performance computing joint undertaking known as EuroHPC.4 Sweden's AI research centres, anchored by institutions such as the Wallenberg AI, Autonomous Systems and Software Program, produce work that competes at the global frontier. Danish research in natural language processing, healthcare AI, and climate modelling contributes disproportionately to European AI scholarship relative to the country's size.
What the Nordics lack
The Nordic countries lack three things that no amount of governance quality can substitute: scale, capital, and control over their own computing infrastructure.
Scale is the most fundamental constraint. The combined population of Denmark, Finland, Iceland, Norway, and Sweden is approaching 28 million, smaller than many individual metropolitan areas in the United States, China, or India. Combined economic output is roughly $1.8 trillion, substantial per head and modest in aggregate. A single American technology company such as Apple, Microsoft, or NVIDIA carries a market capitalisation exceeding the combined economic output of all five countries.
The talent pool reflects the same constraint. Nordic universities produce excellent AI researchers in modest numbers. Competition for AI talent is intense, and the salaries, equity packages, and research budgets available at American and Chinese technology companies routinely draw the strongest researchers away from European institutions. Sweden and Finland have partly addressed this through immigration policy and international research collaboration, but the net flow of frontier AI talent runs away from the Nordics rather than toward them.
Capital is the second constraint. Nordic venture investment has grown, and the ecosystem crossed $500 billion in cumulative startup valuations in late 2025, with quarterly deal flow reaching $1.8 billion in the third quarter, a seven-quarter high. That is nonetheless a fraction of the capital deployed in AI globally. The OECD reports that the United States alone accounted for about 75 per cent of worldwide AI venture capital deal value in 2025, some $194 billion, against 6 per cent for the twenty-seven member states of the European Union combined, or $15.8 billion. The Nordics sit inside that 6 per cent as a smaller fraction still.
The gap is structural rather than merely quantitative. Training a frontier model requires investment of tens or hundreds of millions of dollars in computing alone. Operating a large language model at scale carries ongoing infrastructure costs exceeding the annual budgets of most Nordic AI research programmes. No single Nordic country can sustain a frontier laboratory at the scale of OpenAI, Anthropic, Google DeepMind, or DeepSeek. The economics of frontier development assume a concentration of capital that the Nordic model, with its distributed governance, moderate taxation, and cautious investment culture, was never designed to provide.
Control over computing infrastructure is the third constraint, and the one tied most directly to the geopolitics of AI. Access to graphics processing unit clusters, data centres, and high-performance computing facilities is a prerequisite for both development and deployment. The EU's AI Giga-Factory initiative addresses this at European level, but the four planned facilities serve the entire continent. Nordic participation in European computing infrastructure is real and proportional: LUMI is a European asset sited in Finland rather than a Finnish one, and access is allocated through the EuroHPC framework rather than by national priority.5
The New Nordics AI initiative
New Nordics AI, funded by the Council of Ministers and launched in October 2025 with its base in Stockholm, is the most concrete expression of a Nordic AI strategy discussed in various forms for several years.
Its design reflects the characteristic Nordic approach: cooperative, multi-stakeholder, and focused on governance. The stated objectives are to increase AI adoption across the region, promote joint investment, and strengthen international competitiveness.6 Staff are distributed across all participating countries. The initiative explicitly includes Estonia, Latvia, and Lithuania, extending the framework beyond the traditional Nordic five.
The funding model is layered. The Council of Ministers provides the core 30 million kroner over three years, with additional funding from corporate partners including Google and Microsoft, and from Nordic Innovation, the Council's innovation agency.7 Corporate participation brings resources and expertise, and it also means the centre's priorities will be shaped in part by the strategic interests of companies whose AI platforms the centre exists to complement or challenge.
Computer Weekly noted both the ambition and the mismatch of scale: five nations pooling resources to compete in a domain where individual companies deploy more capital than the whole initiative's multi-year budget.8 The open question is whether the centre functions as a genuine coordination mechanism, pooling computing capacity, aligning research priorities, and harmonising governance, or as a gesture allowing ministers to announce action without providing the resources to sustain it.
The early signs are mixed. AI Sweden is well established and has developed a model for public-private collaboration regarded as among the most effective in Europe. AI Finland contributed to Elements of AI, a course that has taught AI fundamentals to people in more than a hundred countries. Digital Dogme has built a community of Danish organisations committed to digital responsibility. These are not negligible assets, but they are assets for adoption and governance rather than for development at the frontier. New Nordics AI can make the region a leader in responsible AI use. At its current funding level it cannot make the region a leader in AI capability.
Sovereignty at small scale
The Nordic paradox is a specific instance of a broader challenge: how small, well-governed states maintain meaningful sovereignty in a domain where power concentrates in entities that dwarf them in capital, computing capacity, and market reach.
The Berlin summit on European digital sovereignty on 18 November 2025 brought the challenge into focus, producing twelve billion euros in voluntary corporate commitments alongside the European Commission's separate twenty billion euro commitment to AI Giga-Factories, the large-scale facilities intended to anchor European AI development.9 The Nordic countries participate in these European initiatives, but their voice within them is proportional to their size. Finland's contributions to the Paris AI Action Summit, Denmark's advocacy for ethical AI standards within the Council of the European Union, and Sweden's role in standardisation bodies are all meaningful, and all operate within frameworks designed for a continent rather than a region.
The tension runs between two models of sovereignty. The European model seeks scale through aggregation, pooling the resources of twenty-seven member states to compete with the United States and China. The Nordic model seeks influence through quality, demonstrating that well-governed, high-trust societies can develop and deploy AI more responsibly, equitably, and sustainably than larger but less well-governed actors.
These models are not mutually exclusive, and the Nordics pursue their regional strategy while participating actively in European initiatives. But the European model subordinates regional priorities to continental ones. The Giga-Factories will be sited where they make most economic and infrastructural sense, which may or may not be in the Nordic countries. European computing capacity is allocated through frameworks reflecting European-wide priorities. The Nordics can contribute to European AI sovereignty. Whether they can maintain a specifically Nordic sovereignty, meaning the ability to develop and deploy AI according to Nordic values and priorities, is a different and harder question.
Sferical AI, launched in August 2025 by AstraZeneca, Ericsson, Saab, SEB, and Wallenberg Investments, represents an alternative approach: sovereign computing infrastructure built by Nordic companies for Nordic needs. Based in Linköping and built on two NVIDIA DGX SuperPOD installations combining 1,152 interconnected processors, it aims to give Swedish industry access to high-end computing on secure, sovereign infrastructure. Whether a single national initiative can achieve what continental programmes have struggled to deliver remains to be seen. The impulse behind it, to build infrastructure under Nordic control rather than relying on American hyperscale providers or shared European facilities, reflects a recognition that sovereignty requires physical infrastructure and not only governance frameworks.
The governance advantage
The argument that governance quality is itself a competitive advantage, rather than merely a constraint on development, has been made frequently by Nordic policymakers and is not without merit.
Trust matters for AI. Citizens who trust their governments are more willing to share data, use digital services, and accept AI-augmented public services. Organisations operating within predictable, transparent regulation can invest in compliance with more confidence than those navigating opaque or arbitrary governance. Researchers working under strong ethical oversight produce work the international scientific community accepts more readily. These advantages are real and they compound.
The adoption figures bear this out. Eurostat records that in 2025 Denmark led the European Union with 42 per cent of enterprises using AI technologies, followed by Finland at 37.8 per cent and Sweden at 35 per cent, against an EU average of 20 per cent. Denmark and Finland also recorded the largest year-on-year increases in the Union, at 14.5 and 13.5 percentage points. Nordic governance is not hindering AI adoption. It is facilitating it, at least within the bounds of what is available to adopt.
The counter-argument is equally clear: governance advantages are advantages only if there is something to govern. The best regulatory framework in the world cannot produce a frontier model. The most trusted digital identity system cannot compensate for the absence of a domestic large language model trained on Nordic languages and cultural contexts. The most transparent data governance cannot substitute for access to the computing infrastructure needed to train and deploy at scale.
The risk is that the Nordic countries become excellent consumers of AI developed elsewhere, deploying American and Chinese models within Nordic governance frameworks, adapting global technology to local needs, and setting standards for responsible use, without developing the autonomous capability to shape the technology itself. This is not hypothetical. The overwhelming majority of AI models used in the Nordic countries are built by non-Nordic companies. The governance advantage applies to how AI is used, not to what AI is built.
Small states and concentrated capability
The Nordic paradox is not unique. It is the challenge facing every small, well-governed country in a world where AI capability concentrates in a handful of entities operating at scales no individual state, and few collections of states, can match. Switzerland, Singapore, Israel, the Netherlands, and South Korea all face versions of the same dilemma.
The Nordic response, pooling sovereignty, sharing infrastructure, and coordinating governance, is the most developed institutional attempt to address it. It builds on decades of Nordic cooperation in defence, trade, education, and research. It draws on genuine regional strengths in digital governance, institutional trust, and public-sector innovation. And it repeats at regional scale the strategy the European Union pursues at continental scale: compensating for the absence of individual hyperscale capability through collective action.
Whether the strategy succeeds depends on what success means. If it means building a Nordic frontier laboratory to compete with OpenAI, Google DeepMind, or DeepSeek, the answer is almost certainly no, because the capital, computing, and talent requirements exceed what the region can mobilise even collectively. If it means ensuring that AI developed anywhere can be deployed responsibly, equitably, and transparently within the Nordic countries, and that the Nordic model serves as an international benchmark for how AI should be governed, the answer is more hopeful.
The harder question is whether governance without capability is sustainable. In a world where AI increasingly mediates access to information, shapes economic opportunity, influences healthcare and education, and informs decisions about security and defence, a region that governs AI well without developing it autonomously is dependent—on the companies and countries that do build the technology, and on their willingness to supply it on terms that respect Nordic values.
That dependence is the core of the paradox. The Nordic countries built everything responsible AI requires except the AI itself. They have the institutions, the trust, the digital literacy, the regulatory frameworks, and the democratic accountability. What they do not have, and what four million euros and a joint ministerial declaration cannot provide, is the scale of capital, computing, and market that this era of AI development demands.
The question is not whether the Nordic model is good. It is good, measurably and demonstrably and in some respects uniquely so. The question is whether good governance is enough in a domain where power follows capital and capital follows scale. The answer will define not just the Nordic countries' place in the AI era, but whether responsible AI is a viable national strategy or a luxury left to those who cannot afford the alternative.
Footnotes
-
AI Sweden, "Five Nordic Ministers Agree: The Nordic Region Must Be Europe's Driving Force in the AI Race," 2025. The declaration is widely reported in these terms; the wording here follows that reporting rather than a retrieved copy of the declaration itself. ↩
-
Finland's Ministry of Economic Affairs and Employment, "Finland's Age of Artificial Intelligence," 2017. ↩
-
ScienceDirect, "Trust, Transparency, and Openness: How Inclusion of Cultural Values Shapes Nordic National Public Policy Strategies for Artificial Intelligence," 2020. ↩
-
EuroHPC Joint Undertaking, "LUMI—Large Unified Modern Infrastructure," Kajaani, Finland. ↩
-
EuroHPC Joint Undertaking, access allocation framework, 2025. ↩
-
Nordic Council of Ministers, "Nordic Council of Ministers Approve Funding for a Nordic-Baltic AI Center," June 2025. ↩
-
AI Sweden, "The Nordic Council of Ministers Grants Nordic AI Center 30 Million DKK," 2025. ↩
-
Computer Weekly, "Nordic States Launch Ambitious AI Centre Regional Plan," 2025. ↩
-
German Federal Ministry for Digital and State Modernisation, "Summit for More Digital Sovereignty Starts in Berlin," 18 November 2025. The twelve billion euro figure is confirmed in summit reporting; the twenty billion euro Giga-Factory commitment is a separate European Commission pledge rather than a summit outcome. ↩